How accurate are self-reported cardiovascular disease events compared to medical record adjudication and hospital morbidity databases?
Self-reported CVD events have a moderate verification rate (68.1%) by medical records, but combining self-report with hospital database linkage improves event identification in epidemiological studies.
Epidemiological studies often rely on self-reported cardiovascular disease (CVD) information, but this may be inaccurate. We investigated the accuracy of self-reported CVD (myocardial infarction, stroke, coronary artery bypass surgery and coronary artery angioplasty) during the follow up of the Australian Diabetes, Obesity and Lifestyle (AusDiab) study. Self-reported CVD events, including the date of the event and hospital admission details, were collected with an interviewer-administered questionnaire. Of the 276 self-reported CVD events, 188 (68.1%) were verified by adjudication of medical records. Furthermore, linkage to the statewide Western Australian Hospital Morbidity Database (WAHMD) showed that CVD events were unlikely to be missed, with only 0.2% of those denying any CVD event being recorded as having had an event on the WAHMD. The adjudication of medical records was as accurate as record linkage to the WAHMD for validation of self-reported CVD, but combining the results from both methods of ascertainment improved CVD event identification.
Barr et al. (2009) studied this question.